Linked Data causes ~25% overhead/DB explosion

[tdb01r]

The rdf field (when populated) adds ~58 values per record, corresponding to ~300 DB rows.

• any dependent field requires re-writing the entire rdf field (300 rows+)
• requesting an eprint reads all rows

In testing this adds ~25% overhead on all EPrints retrieval-based operations.

For a medium-sized repository (10k records) the rdf tables will be 500k rows long. This is not scalable for a core field.

[cjg]

The RDF does *not* need to be retrieved when doing $dataset->dataobj(xxx) etc.

Would it make more sense to store the RDF in a stand alone table (or dataset?)

[cjg]

Fixed in r5090. Now stored in a separate dataset.